Running is one of the most popular exercises to maintain physical and mental health. However, high speeds during exercise may introduce excessive impact forces and overstriding and can lead to increased injury risks. Recent advancements in machine learning and wearable motion systems are promising approaches for in real-time monitoring of running performance and injury prevention. The main objective of our research project is to evaluate the contribution of the ground reaction force (GRF) measurements in monitoring running speed, using efficient and reliable features and machine learning models to accurately classify athletes’ running speed levels. Based on Fukuchi et al. (2017) public dataset in which running performances was collected for 28 athletes with 3 different speeds, 10 features from GRF were extracted 28 athletes to represent relevant biomechanical determinants of running. Four classification algorithms were tested (LD, KNN, SVM, RF) to classify the 3-speed levels (Low, moderate, and High) and evaluated through 2 criteria (accuracy, execution time). Results from statistical tests and correlation demonstrated that only 7 features, independent and less correlated ones, from vertical and anterior posterior GRF are pertinent for classification. Based on performance criteria for different tested classifiers, KNN outperforms others when combining features from AP and vertical forces and demonstrates the shortest execution time. The results provide an insight into potential advances in running analysis based on artificial intelligence and speed monitoring to enhance athlete performance during training and competition.